【作者】 叶晖；

【导师】 陈志刚；

【作者基本信息】 中南大学 ， 计算机应用技术， 2010， 博士

【摘要】 机会网络是一种不需要源节点和目标节点之间存在完整的通信链路,利用节点移动带来的相遇机会实现数据通信的自组织网络。作为一种全新的组网方式,它突破了传统网络对于延迟和传输率的限制,使得机会网络在很多领域有着巨大的应用潜力。其数据通信有着如下典型特征：1)节点移动频繁,网络常处于不连通状态,数据端到端延时较长。2)数据传输强依赖于节点移动带来的相遇机会,从而以刻画节点相遇概率与时间周期分布的移动模型成为影响网络性能的关键因素。3)缓存空间需求矛盾比MANET更加突出；由于数据在网络中要滞留较长时间,使得中继节点的存储空间消耗较快。其对缓存容量有着更高的要求。机会网络数据分发的研究难点在于如何在高延迟、低数据率的网络环境下确保其效率与数据的可靠传输。由于机会网络节点采用“缓存-移动-转发”的机制进行数据通信,这就需要对于协作缓存、缓存替换、路由以及支持移动性的底层通信协议等方面进行全面地考虑。基于此,本文在提高数据分发效率和传输性能密切相关的算法和协议方面展开研究,主要的工作和创新性成果包括：(1)基于刻画相遇概率与时间周期分布的机会网络节点移动模型,提出了一种机会网络节点协作缓存优化策略已有缓存策略研究并未将节点移动规律作为一个重要因素进行考虑,实际上在机会网络中节点的移动规律影响着节点之间的协作关系与局部域的稳定,并进一步影响着数据块的有效性和可用性。为此,提出了一种以人为载体的机会网络协作缓存优化策略：HMP-Cache。深入研究了人类移动模式中节点运动状态与相对静止状态下的特点,利用目标地址匹配标准来严格选择协作缓存节点,对数据无用多跳转发造成的影响进行有效控制；并采用同步Cache数据表来达到局部域内缓存信息共享的目的,力求弥补单个节点缓存资源有限的不足,从而进行cache信息有效共享,以此达到充分利用各个移动节点有限的缓存资源,有效提高热点数据命中率的目标。(2)基于前述的协作缓存优化策略,以节点与数据相关度为标准,进一步提出了一种机会网络节点缓存替换策略已有缓存替换策略工作大多是以数据的相关访问信息作为替换标准,如访问次数,最后一次访问时间,数据项尺寸大小等,以此来对数据项替换策略进行设计。而对于数据的有效性考虑不够。为此,提出了一种新的节点缓存替换策略：ON-CRP,深入讨论了节点移动模式对缓存区内数据项“有用性”的影响,采用特定数据与特定节点的相关度来进行替换标准的设计；同时将数据的访问与更新频率作为影响缓存数据有效性的重要因素进行了设计；并讨论了齐夫分布在不同的参数取值情况下对性能的影响。仿真实验结果表明本文策略能够有效降低数据的远程访问延迟,提高数据缓存命中率。(3)根据节点不同运动状态,利用多因素综合评判理论,提出了一种采用混合数据转发行为的机会网络路由算法目前已有的机会网络路由研究绝大多数仅采用单一的数据转发行为来进行设计,并未结合节点运动状态的改变来综合考虑性能。同时,单一的数据转发行为往往并不能完全适用于现实环境。针对现有工作的不足,本文根据节点不同运动状态进行相应的数据转发行为设计,在节点正常移动状态下,采用节点信任度,目标地址匹配标志,网络跳数来严格选择有效的中继节点以控制无用多跳转发的影响和资源耗费,在节点小幅度往返移动的状态下则采用有限洪泛来降低数据时延。在此基础上提出了一种采用混合数据转发行为的机会网络路由算法,并且讨论了算法的性能上下界。(4)基于无比率编码理论,提出了一种改进移动状态下数据传输性能的机会网络MAC层通信协议由于机会网络节点的移动特性使得目前应用程度较高的蓝牙协议在数据通信过程中丢包率,抖动与随机链路错误有所上升。为此,本文提出了一种基于BlueTorrent改进的协议：BlueTorrent-ED,协议对蓝牙的关键通讯参数集进行优化设计,并采用无比率编码算法进行了数据分发优化,从而降低数据丢包影响并提高网络可靠性。更多还原

【Abstract】 Opportunistic network is self-organization Network that utilizes the mobile node contact opportunity to achieve data communications. It does not require one complete communication path exists between source and destination. As a new Network formation, it breaks through the data latency and transfer rate restrictions of traditional networks. Therefore, Opportunistic network has great potential applications in many areas. It is more suitable for practical self-organization network applications than traditional Mobile Ad Hoc Network in real environment. There are the following typical characteristics in data communication of opportunistic networks.1) The nodes move frequently with large range in opportunistic networks, the network is often divided into several sub-regions which are not mutually connected. Moreover, the low wireless channel capacity leads to the long data end to end latency.2) The data communication depends on contact chance arising from node movement. Thus the node mobile model that characterizes the probability of the node contact and time period data distribution becomes the key factor in network communication performance.3) Contradictions of node cache space are more prominent than the MANET, data lives long time in network because relay nodes adopt "store-carry-forward" pattern to transfer data. Therefore, relay node’s cache space consume fast. It has a higher buffer capacity requirements comparing with traditional MANET.Difficulty of Opportunistic network data dissemination is to ensure its efficiency and reliability of data transmission in high latency and low data rate network environment. It needs to think about cooperative cache, cache replacement, routing and MAC layer protocol and so on because nodes adopt "cache-move-forward" mode to communication. Therefore, the main work and contributions are presented in the following aspects:(1) An Opportunistic network data cooperative caching policy based on mobile Patterns of Humans is proposed, which is characterized by the probability of the node contact and time period data distributionThe node movement pattern is not considered as an important factor in many of existing policies that has impact on cache performance. In fact, node movement trajectory affects collaborative relationships between nodes and the stability of the local network domain. Moreover, it has impact on the effectiveness and availability of data blocks. Therefore, a cooperative caching policy based on human mobility patterns is proposed, which is referred as HMP-Cache. HMP-Cache uses the standard of target address matching to choose cooperative caching nodes. Also, the sharing caching information is done by synchronization of caching table in local region. Therefore, the shortcoming of limited caching resources is compensated and the remote data latency is reduced. Moreover, the cache hit ratio is improved.(2) Taking the correlation between the nodes and the data item as replace standard, a cache replacement policy for opportunistic networks is further proposed, which is based on the foregoing cooperative caching policy.Most existing cache replace policies take the data access information as replace standard, such as access times, the data item size and the last access time. But the validity of data is not considered in existing research. Thus, a novel cache replacement policy is proposed, which referred as ON-CRP. ON-CRP chooses the data item to replace based on the correlation between the nodes and the item. It utilizes the probability of destination-matching to judge the correlation which is gotten from humans mobile patterns. Furthmore, the visit and update frequency ratio is used to design the standard of cache data replacement. In addition, the performance impact of Zipf distribution parameter is dicussed. The simulation results show that our policy can reduce the latency of remote data access while cache hit ratio of data is improved.(3) Using a general evaluating mechanism, an opportunistic networks routing policy based on node different movement staus is proposed, which uses hybird data forwarding behaviors.Most existing routing algorithms assume a single forward behavior during the entire data transfer process without considering human movement changes in mobility patterns. In the real world, however, a single forward behavior may not be applicable in different contexts. Therefore, we utilize different data forwarding behavior according to the node move staus. The key factor of data dissemination policy is discussed in depth. In normal move status, the network traffic and resource consumption caused by useless data dissemination of multiple hops are reduced efficiently through the design of selecting relaying nodes. In small move status, the limit flooding is used to reduce the data latency. Furthermore, a routing policy based on node different movement staus is proposed referred as HMP. Moreover the upper and lower bounds of HMP performance are analyzed.(4) A Bluetooth MAC layer protocol based on rateless coding theory is proposed, which is improved data transmission performance upon bluetorrent protocol.Bluetooth protocol is employed as the MAC layer protocol given its popularity and technical maturity in mobile devices. However, the Bluetooth is a low data rate and short distance communication protocol. The opportunistic networks mobile features make the packet loss rate, jitter and random link error increased in communication. Therefore, a novel Bluetooth protocol is proposed. It is improved upon Bluetorrent, referred as Bluetorrent-ED. Bluetorrent-ED optimizes Bluetooth communication parameters to improve the efficiency. Furthermore, to improve network reliability, On-Line coding algorithm is used for reduce the impact of data drop and optimization in data dissemination.更多还原